Methodology

Components

• Social growth velocity (30%, measured): growth of matching X posts over 1h, 6h and 24h. Short windows are compared with the same hours on previous days, so the normal daily rhythm does not look like growth.
• Cross-platform adoption (20%, measured): how many monitored sources show real activity. With only one monitored source this component is missing.
• Meme potential (20%, AI estimate): an AI estimate with written reasoning and cited posts.
• Originality (15%, AI estimate): an AI estimate with written reasoning and cited posts.
• Low on-chain saturation (15%, measured): 100 minus the saturation index below.

Total = sum of weight × component. An AI model never chooses the total.

Saturation index

Solana tokens matching a narrative are counted once per token and weighted by match confidence (verified 1.0, probable 0.6, possible 0.2) and by trading activity (active 1.0, low 0.6, inactive 0.15). Inactive means total liquidity under $1,000 or almost no trading in 24 hours, so dead tokens barely count. Index = 100 × (1 − e^−W/2.5). A failed search gives an unknown index, never “no tokens”.

Missing data

• A missing component counts as 0 and the score is marked provisional (a lower bound).
• Without growth data, or with less than 50% of the weight available, no score is shown.
• Stale data is labeled, and growth is not calculated from incomplete windows.

Automatic narrative detection

Candidate terms come from words shared by several new token names and from hashtags on shared TikTok videos. An AI filter removes generic words, politics, sports teams and countries. Remaining candidates are checked against X post counts and are added only when they are clearly rising versus their usual level.

Limits

• Counts reflect only the sources listed on the About page. Many conversations happen elsewhere.
• Token matching uses names and symbols and can miss tokens or include unrelated ones.
• Observed momentum often fades. A high score says nothing about a token's future price.